If you’ve ever used Heroku, you’re probably familiars with its logs. Logs a great when debugging a test app as they are readable enough while providing useful information. However, once the application is recieving a lot of traffic, it become quite difficult to make sense of all the logs!
There’s a solution for that, using angle-grinder. Here are some sample commands to use. They can easily be tweaked.
- Inspired by Power tools for analyzing your Heroku logs
- Heroku logs documentation
angle-grinderon Github
Status Aggregate by Second #
Command #
heroku logs -t | agrind \
'*
| parse "T*." as time
| logfmt
| count(status < 400) as successful,
count(status >= 400) as failure,
p95(service) as service,
p95(connect) as connect
by time
| sort by time desc'
Sample
time successful failure service connect
-------------------------------------------------------------------------------
09:24:18 9 0 150 0
09:24:17 11 0 29 0
09:24:16 33 0 106 0
All Paths #
Command #
heroku logs -t | agrind \
'path | logfmt' \
--format '{dyno:<6} {connect:6} {service:7} {status:4} {method:5} {path}'
Sample
web.1 0ms 5ms 304 GET /foo?bar=baz
web.2 0ms 32ms 200 GET /hello/word?user=1337
web.3 1ms 120ms 400 POST /auth/attempt
All status >= 400 #
Command #
heroku logs -t | agrind \
'path | logfmt | where status >= 400' \
--format '{dyno:<6} {connect:6} {service:7} {status:4} {method:5} {path}'
Sample
web.3 0ms 8ms 404 GET /not/found
web.2 1ms 25ms 500 GET /server?error
web.1 0ms 221ms 403 POST /forbidden/path
First paths stats #
Command #
heroku logs -t | agrind \
' path
| parse regex "path=\"(?P<canonical>/[^/?\"]+)(?P<rest>[^\"]*)?\"" nodrop
| logfmt
| count(status < 400) as successful,
count(status >= 400) as failure,
p50(service) as mean,
p95(service) as p95,
p99(service) as p99
by canonical
| sort by mean desc'
Sample
canonical successful failure mean p95 p99
----------------------------------------------------------------------------------------------
/base-path 123 1 61 93 61
/topLevel 15 0 39 83 103
/api. 17 0 33 67 70
